FIG. 1 illustrates a method for transmitting audio signals in today's communication devices. As shown in FIG. 1, the method includes first performing voice activity detection to determine whether the current audio frame contains speech or noise. Voice activity detection typically includes a signal feature extraction module 11 and a speech/noise decision module 12 as shown in FIG. 2. In the signal feature extraction method module 11, feature vectors of the current frame are extracted. With these feature vectors, the speech/noise decision module 12 decides whether the current frame contains noise or speech. The reason for distinguishing speech from noise using voice activity detection is because typical audio sequences contain a lot of noise (e.g., sometimes approaching 50% of the signal). Thus, coding/decoding the speech and noise using the same method can be wasteful and unreasonable. Accordingly, coding/decoding speech and noise differently after distinguishing them would be desirable to, for example, reduce the number of bits and the amount of coding/decoding calculation.